I don't think it's entirely disconnected. It'd be disconnected if you had samples that are comparable (ideally random, representative of the population), but t

I don't think it's entirely disconnected. It'd be disconnected if you had samples that are comparable (ideally random, representative of the population), but that's not the case in most countries. When you start with very limited testing capacity, you prioritize "important" groups of people that are often at higher risk of infection (healthcare workers, older people, ...) so you can expect higher percentage of people to test positive. But as you increase the capacity / number of tests, the sample structure will change - people with lower risk of infection will get tested, so it's natural to expect the percentage positive to decrease.